1. Field of the Invention
The present invention relates to a storage system which can enhance scalability and fault tolerance.
2. Description of the Related Art
In the current data storage system market, very large scale storage systems to which a plurality of host computers are connected are demanded. For example, for the use in a data center of a large company or the like, the large scale storage system having high performance and reliability are demanded.
On the other hand, with the spread of the open systems market, demands for small scale storage systems which have the same performance and reliability as those of the large scale storage systems are increasing.
In order to meet these demands, it is preferable to implement storage systems having high scalability such that a small scale configuration and a very large scale configuration are supported by a common architecture. For this reason, various configurations which implement an arbitrary scale storage system by coupling the arbitrary number of host interface control units for controlling the connections to host computers with the arbitrary number of drive control units for controlling data access to disk drives has been suggested.
A crossbar switching network has a configuration in which a plurality of host interface control units and a plurality of drive control units are coupled with one another via one switch (for example, see Japanese Patent Laid-Open No. 2000-99281).
A cluster switching network has a configuration in which a plurality of clusters are coupled with one another. Here, the cluster means that the plurality of host interface control units and the plurality of drive control units are coupled with one another via one switch (for example, see Japanese Patent Laid-Open No. 2003-99206).
A completely-connected network has a configuration in which the plurality of host interface control units and the plurality of drive control units are all coupled with one another via switches which are provided in the respective control units (for example, see “EMC Symmetrix DMX Series”, [online], EMC Corporation Homepage, Products, [search on Jan. 7, 2005], Internet <URL: http://www.emc.com/products/systems/symmetrix/DMX_series/p df/DMX_series.pdf>
According to the crossbar switching network, the scale of the storage system to be constructed is limited by the scale of the switch. That is, in order to construct the large scale storage system, it is necessary to use a large scale switch. If the large scale switch is used in the small scale storage system, the cost of the small scale storage system increases.
The cluster switching network is intended to relieve the problem of the cost increase by the large scale switch in the crossbar switching network. That is, the cluster is constructed by using a small scale switch. Then, the large scale system which requires for the large scale switch is implemented by connecting the plurality of clusters to one another. However, in order to suppress the cost increase of the switch, there is no choice but to limit data transmission capability among the clusters. For this reason, when data is transmitted over the clusters, performance is significantly degraded, as compared to a case in which data is transmitted within the cluster. As such, in the cluster switching network, performance ununiformity occurs according to a data transmission path.
According to the completely-connected network, when the large scale storage system is constructed, lots of the connections among the switches are required, such that the configuration is complicated. As a result, the cost of wiring lines increases and the switch becomes large.